Marius Mutu
d175d5ba5a
Pre-existing uncommitted changes swept in with the STT work: anaf-monitor snapshots/versions, cron job + newsletter state, 9 youtube KB notes, tools/ocr_bon.py, and tools/tts.py. Note: the tts.py change breaks 2 truncation tests in test_voice_normalize.py (sanitize word-count) — flagged for a separate follow-up. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
133 lines
4.2 KiB
Python
133 lines
4.2 KiB
Python
#!/usr/bin/env python3
|
||
"""Text-to-speech via Supertonic local server.
|
||
|
||
CLI:
|
||
python3 tools/tts.py --text "Salut Marius" [--voice M1] [--lang ro]
|
||
→ stdout: {"ok": true, "path": "/tmp/echo-tts-xxx.wav", "size_bytes": 12345}
|
||
→ stdout: {"ok": false, "error": "..."}
|
||
|
||
Module:
|
||
from tools.tts import synthesize
|
||
result = synthesize("text", voice="M1", lang="ro")
|
||
"""
|
||
|
||
import argparse
|
||
import json
|
||
import sys
|
||
import tempfile
|
||
|
||
import httpx
|
||
|
||
SUPERTONIC_URL = "http://127.0.0.1:7788"
|
||
VOICES = {"M1", "M2", "M3", "M4", "M5", "F1", "F2", "F3", "F4", "F5"}
|
||
DEFAULT_VOICE = "M2"
|
||
DEFAULT_LANG = "ro"
|
||
|
||
# Punctuation Supertonic synthesis rejects with HTTP 500 (Romanian curly quotes,
|
||
# smart dashes, ellipsis, angle quotes). Mapped to ASCII so a stray „foo" in
|
||
# any caller's text doesn't kill the whole request.
|
||
_TTS_PUNCT_MAP = {
|
||
'„': '"', '“': '"', '”': '"',
|
||
'‘': "'", '’': "'", '‚': "'",
|
||
'«': '"', '»': '"',
|
||
'–': '-', '—': '-',
|
||
'…': '...',
|
||
}
|
||
|
||
|
||
# Supertonic ONNX model hard limit: inputs longer than this trigger
|
||
# Mul node dimension mismatches in attention layers.
|
||
_MAX_TTS_CHARS = 400
|
||
|
||
|
||
def sanitize_for_supertonic(text: str) -> str:
|
||
"""Replace Unicode punctuation and strip chars that crash Supertonic's ONNX model."""
|
||
for src, dst in _TTS_PUNCT_MAP.items():
|
||
text = text.replace(src, dst)
|
||
# Strip emoji and high-codepoint chars (keep ASCII printable + Latin/Romanian diacritice)
|
||
cleaned = []
|
||
for ch in text:
|
||
cp = ord(ch)
|
||
if (32 <= cp <= 126) or (128 <= cp <= 591):
|
||
cleaned.append(ch)
|
||
else:
|
||
cleaned.append(' ')
|
||
text = ' '.join(''.join(cleaned).split())
|
||
if len(text) > _MAX_TTS_CHARS:
|
||
text = text[:_MAX_TTS_CHARS]
|
||
return text
|
||
|
||
|
||
def synthesize(text: str, voice: str = DEFAULT_VOICE, lang: str = DEFAULT_LANG) -> dict:
|
||
"""Call Supertonic server and save audio to a temp WAV file.
|
||
|
||
Returns:
|
||
{"ok": True, "path": "/tmp/echo-tts-xxx.wav", "size_bytes": N}
|
||
{"ok": False, "error": "mesaj eroare"}
|
||
"""
|
||
if not text or not text.strip():
|
||
return {"ok": False, "error": "Text gol."}
|
||
|
||
text = sanitize_for_supertonic(text)
|
||
|
||
voice = voice.upper()
|
||
if voice not in VOICES:
|
||
voice = DEFAULT_VOICE
|
||
|
||
try:
|
||
resp = httpx.post(
|
||
f"{SUPERTONIC_URL}/v1/audio/speech",
|
||
json={
|
||
"model": "supertonic-3",
|
||
"input": text,
|
||
"voice": voice,
|
||
"response_format": "wav",
|
||
"lang": lang,
|
||
},
|
||
timeout=60.0,
|
||
)
|
||
resp.raise_for_status()
|
||
except httpx.ConnectError:
|
||
return {
|
||
"ok": False,
|
||
"error": (
|
||
"Serverul Supertonic nu rulează pe :7788. "
|
||
"Pornește cu: systemctl --user start supertonic-tts"
|
||
),
|
||
}
|
||
except httpx.HTTPStatusError as e:
|
||
body = e.response.text[:300]
|
||
# Fallback: dacă lang=ro eșuează, încearcă na (language-agnostic)
|
||
if lang != "na":
|
||
return synthesize(text, voice=voice, lang="na")
|
||
return {"ok": False, "error": f"HTTP {e.response.status_code}: {body}"}
|
||
except Exception as e:
|
||
return {"ok": False, "error": str(e)}
|
||
|
||
# Salvează în fișier temp
|
||
try:
|
||
fd, path = tempfile.mkstemp(prefix="echo-tts-", suffix=".wav")
|
||
with open(fd, "wb") as f:
|
||
f.write(resp.content)
|
||
return {"ok": True, "path": path, "size_bytes": len(resp.content)}
|
||
except Exception as e:
|
||
return {"ok": False, "error": f"Scriere fișier: {e}"}
|
||
|
||
|
||
if __name__ == "__main__":
|
||
parser = argparse.ArgumentParser(description="Supertonic TTS CLI")
|
||
parser.add_argument("--text", required=True, help="Text de convertit în audio")
|
||
parser.add_argument(
|
||
"--voice", default=DEFAULT_VOICE,
|
||
help="Voce: M1-M5 (masculin) sau F1-F5 (feminin). Default: M1"
|
||
)
|
||
parser.add_argument(
|
||
"--lang", default=DEFAULT_LANG,
|
||
help="Limbă (ro, en, na). Default: ro. Fallback automat la na dacă ro eșuează."
|
||
)
|
||
args = parser.parse_args()
|
||
|
||
result = synthesize(args.text, voice=args.voice, lang=args.lang)
|
||
print(json.dumps(result, ensure_ascii=False))
|
||
sys.exit(0 if result.get("ok") else 1)
|